Re: [PATCH v2 10/17] linux-user/aarch64: Implement PR_MTE_TCF and PR_MTE_TAG

[Richard Henderson <richard.henderson@linaro.org>]

On 6/25/20 11:50 AM, Peter Maydell wrote:
> On Fri, 5 Jun 2020 at 05:17, Richard Henderson
> <richard.henderson@linaro.org> wrote:
>>
>> These prctl fields are required for the function of MTE.
>>
>> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
> 
> This API doesn't seem to be in the upstream kernel yet, right?

It has finally landed in 5.10.
Thus the long delay since my last revision.

>> +                    /*
>> +                     * Write PR_MTE_TCF to SCTLR_EL1[TCF0].
>> +                     * Note that the syscall values are consistent with hw.
>> +                     */
>> +                    env->cp15.sctlr_el[1] =
>> +                        deposit64(env->cp15.sctlr_el[1], 38, 2,
>> +                                  arg2 >> TARGET_PR_MTE_TCF_SHIFT);
> 
> This actually will be per-thread since each linux-user thread has
> its own CPU. You probably need to do something to make it be
> inherited across clone and fork, though (assuming those are
> the required semantics).

FWIW, these are a per-thread quantities.

>From arch/arm64/kernel/mte.c:
> void mte_thread_switch(struct task_struct *next)
> {
>         if (!system_supports_mte())
>                 return;
> 
>         /* avoid expensive SCTLR_EL1 accesses if no change */
>         if (current->thread.sctlr_tcf0 != next->thread.sctlr_tcf0)
>                 update_sctlr_el1_tcf0(next->thread.sctlr_tcf0);
>         update_gcr_el1_excl(next->thread.gcr_user_incl);
> }

I don't think I have to do anything special wrt fork/clone, as env->cp15 will
get copied by our own cpu_copy().


r~